10.7 Extras

Here are some ideas for you to add to these projects.

10.7.1 Hypothesis testing

The computations for mean, variance, standard deviation, and standardized Z-scores involve floating-point values. In some cases, the ordinary truncation errors of float values can introduce significant numeric instability. For the most part, the choice of a proper algorithm can ensure results are useful.

In addition to basic algorithm design, additional testing is sometimes helpful. For numeric algorithms, the Hypothesis package is particularly helpful. See https://hypothesis.readthedocs.io/en/latest/.

Looking specifically at Project 3.5: Standardize data to common codes and ranges, the Approach section suggests a way to compute the variance. This class definition is an excellent example of a design that can be tested effectively by the Hypothesis module to confirm that the results of providing a sequence of three known values produces the expected results for the count, sum, mean, variance...